CS/CMPE 4760: Advanced Computer Architecture, Fall 1998

Professor: Ann Chervenak
Office: 218 College of Computing
Phone: 894-8591
Email: annc@cc.gatech.edu
Office Hours: Mon/Wed 1-2pm
T.A.: Ramkumar Krishnan
Email: ramkumar@cc.gatech.edu
Office Hours: Tues/Thurs 11am-12N
Held in Student Commons Area, first floor of CCB

Project 3 Web Pages


Lecture Notes


Basic Pipelining (Chapter 3): ( postscript or pdf or ascii )

*** UPDATED *** Advanced Pipelining (Complete) (Chapter 4): ( postscript or pdf or ascii )

Notes on Intel/HP Epic: ( postscript or pdf or ascii )

Notes on Caches, Memory and Disks: ( postscript or pdf or ascii )

Notes on Multiprocessors: ( postscript or pdf or ascii )

Exams


Midterm 1: Friday, October 23, 1998
Students may bring calculators and one sheet of notes (2-sides)

Sample Midterms
Note: Earlier classes may have covered material in a different
order, so some questions are not applicable to this class.

Fall 1997 (postscript)
Winter 1997 (postscript)
Fall 1995 (postscript)


Final Exam: Wednesday, December 9, 1998, 8 a.m. to 10:50 a.m.
Students may bring calculators and one sheet of notes (2-sides)

Sample Finals
Fall 1997 Final ( postscript or pdf or ascii )
Fall 1996 Final ( postscript or pdf or ascii )

REVIEW FOR FINAL EXAM ( postscript or pdf or ascii )

Papers


Note: These RAID Papers can print via ghostview but might not be view-able
RAID Paper (short form--required reading)
RAID Paper (long form--includes advanced topics)

Handouts


REVIEW FOR FINAL EXAM ( postscript or pdf or ascii )

Syllabus: ( postscript or ascii )

Photo Page: (Due 9/30/98) ( postscript or ascii )

DLX Architecture: ( postscript or ascii )

Scoreboard Example: postscript

Tomasulo Example: postscript

Blank templates for Scoreboard and Tomasulo Algorithms postscript

Speculative Execution Examples from class:
Example 1: postscript or pdf
Example 2: postscript or pdf


Blank templates for Speculative Execution with Tomasulo postscript or pdf

To help you prepare your presentations and written reports, the sets of slides below
share advice that I've given in the past to graduate students.

ADVICE FOR GOOD TALKS postscript or pdf or ascii

ADVICE FOR WRITING GOOD PAPERS postscript or pdf or ascii

Homework


Homework 1: Due 10/21/98 IN CLASS (no late homework accepted)
( postscript or pdf or ascii )

Homework 2: Due Monday 11/30/98 IN CLASS (no late homework accepted)
( postscript or pdf or ascii )

Projects

Project 3:
Due Friday Nov 6th, in class: project choice and list of partners Specification
For more ideas, look at Wisconsin Computer Architecture Page

Project 2:
Due Tuesday Nov. 24th at 6pm (NO LATE DAYS!!)
Implements Tomasulo's algorithm and speculative execution
Extra credit: implement branch target buffer
Specification

* Note: another slight change in spec * 11/12/98: Changes
Summary of changes:
(1) Put reorder buffer number rather than execution unit number into register result status
(2) For ADDI, ANDI and LW instructions, set Rk=1 and Vk=0 (instead of Rj=1 and Vj=0)

* Note: slight change in spec * 11/4/98:
Put branch target in storeAddress field and branch condition in result field of reorder buffer.

Simulator program template (changed slightly 11/12/98)

* NOTE: Fixed bug on 11/11/98 resulting in minor changes in output files *
Program 1, Machine Code for Program 1, Sample Output for Program 1
Program 2, Machine Code for Program 2, Sample Output for Program 2
Program 3, Machine Code for Program 3, Sample Output for Program 3


EXTRA CREDIT: DYNAMIC BRANCH PREDICTION
Help with dynamic branch prediction

Revised output using branch prediction
Sample Output for Program 1
Sample Output for Program 2
Sample Output for Program 3

Project1:
Specification
Simulator program template
Assembler Executable (compiled on Sun workstation)

New Sample Output for First Project (10/12/98)
Included below is the revised sample output for the first project.
When I changed around my forwarding code, I introduced a couple of errors.
One showed up in the first program in the form of executing the
loop an extra time. The second showed up in the second program
when the forwarding to the andi and addi instructions was incorrect.
These should be fixed now. Sorry for the confusion.
Let me know if you find any other errors.
(Remember, you may need to hit reload if your browser cached the old version.)

Program 1, Machine Code for Program 1, Sample Output for Program 1
Program 2, Machine Code for Program 2, Sample Output for Program 2
Program 3, Machine Code for Program 3, Sample Output for Program 3